The same customer, four times over.
Four fixed prices for customer-record work. The first one changes nothing — it reads what you have and tells you what is actually wrong with it, which is the only honest place to start.
Data Health Check
We read your records and write down what is wrong with them. No changes are made to anything.
$390
Clean and Confirm
The duplicates found, reviewed and merged, with the evidence kept for each decision — and a snapshot of both records before every merge.
$2,950
Move and Confirm
Records moved from one system to another, then checked field by field afterwards. Clean them first if they need it — moving duplicates only moves the problem, and it is harder to unpick once it has landed.
$4,900
There is a cheaper way to do each of these. Here is when to take it.
- Instead of the $390: most CRM firms will audit your records free. Take it if you want a rough idea. Ours is a written report you keep, naming what was examined, what was found, and what could not be settled from the data you have.
- Instead of the $2,950: outsourced teams clean records by hand from about five cents each. That is the right buy when the job is typing. It is the wrong buy when the question is which two records are the same person, because that is a judgement, and ours arrives with the evidence for every decision and a snapshot of both records before any merge.
- Instead of the $4,900: a migration tool moves a hundred thousand records for about $1,199, and charges roughly $875 more to check the move afterwards. If your data is already clean and your fields already line up, buy the tool — we will tell you so. Ours is for the move where they do not line up, and where somebody has to decide what happens to the records that do not fit.
Nobody decides to have duplicate records.
They arrive one at a time. Somebody fills in a form with a work email and again with a personal one. A trade show list gets imported. Two salespeople add the same person on the same afternoon, spelled two ways. None of it is anyone’s fault and all of it adds up.
A business without a programme for this typically runs somewhere between 10% and 30% duplicates. The ones that stay on top of it sit under 2%. The gap between those two numbers is a marketing list that emails the same person twice, a salesperson calling a customer who is already someone else’s, and a report that counts one company as three.
What the $390 check reads.
Which records are the same person.
Compared on the whole record rather than on one field, so a changed email or a married name does not hide a match, and two people at the same company are not merged because they share a domain.
What is missing where it matters.
Not a completeness percentage across every field — the fields that actually stop work when they are empty.
Contact details that cannot be right.
Addresses that no system will accept, numbers with the wrong number of digits, values sitting in the wrong field.
Consent and marketing permissions.
The field that turns a data problem into a regulatory one if it is wrong, and the only one that costs you either way. Lose an opt-out in a merge and you email somebody who asked you not to. Lose an opt-in and a list you are entitled to use quietly stops being usable. We check it in both directions.
Owners and records pointing at people who have left.
Asked of the user list, not of the records — a record still pointing at a deleted user hides from a check that asks the records who owns them.
Nothing is changed at this tier. You get a written report naming what was examined, what was found, and what could not be settled from the data available.
Read a specimen report, start to finish
It is a specimen and says so in its own header, so it still says so if somebody forwards it to you. It was produced by running the procedures against a sandbox we own, filled with invented records. Names and email addresses are replaced by record identifiers, which is what we do for any report shown to anyone other than its addressee.
What the matching scored, and what those numbers do not tell you.
Measured against a set where the right answer was known in advance, and shaped by a live CRM’s own constraints rather than by whatever flattered the result.
100%
Precision, against 400 records that are different people by construction. No false positives, and no group merged two different people. What it does not say: how many real duplicates it never reached.
43.0%
Recall, at entity level. The number to hold us to, and one we have not found any other firm in this category publishing.
Both figures come from the checks as they are built, run against records where every answer was known beforehand. They have not yet been run against a paying customer’s data, and we would rather say that than imply a track record we do not have.
You can watch the matching work on somebody else’s records. Foundation Legal Advisors runs every name on a new enquiry against every name the firm already holds, including former names. The names inside are invented, so try to break it — a typing slip, the same company in two countries, initials instead of a name.
What we will not do.
-
We will not run your systems, only read what they hold.
Everything here reads data that is sitting still. If what worries you is a system that runs on its own — something built by somebody who has gone, that nobody has looked at since — that is a different job. We put real messages through it and force a failure to see what it does. On our published specimen that is how we found a refund of £84 released to somebody who did not own the order. It is a written review, $1,350, and there is a full example free to read before you buy.
-
We will not de-duplicate your company records.
The matching runs on contact records — people. Two accounts for the same business, spelled two ways, is a different job with a different model behind it, and we do not sell it here. If that is the problem you have, say so and we will tell you plainly whether we can help.
-
We will not quote an accuracy figure.
One real run compared 9,901 record pairs, of which four were duplicates. A tool that flagged nothing at all would score 99.96% accurate on it. The UK’s national statistics office removed its own accuracy formula for this reason — it “did not give a good representation of the quality of the linkage” — and recommends always reporting precision and recall instead. So we do.
-
We will not merge a record on the machine’s say-so alone.
Above a high bar it merges. In the band below it, pairs are listed for you to decide on and nothing is touched — on the calibration set that band held 252 pairs, of which 249 were genuine duplicates and 3 were not, and every one of them was left for a person. A merge is not reversible in most systems, which is why the burden of proof sits where it does.
-
We take a copy of both records before we merge them.
Most systems cannot undo a merge. HubSpot has no way to separate two records again once they are joined, and Salesforce’s window for it is short. So before every merge we keep a copy of what both records said. If a decision turns out to be wrong, there is something to go back to — which is not the same as keeping a note of why we did it. Some tools in this category can restore a merge on their higher tiers, one group at a time. We take the copy on every merge, as standard, and it is not a paid extra.
-
We will not check a sample and call it checked.
The normal way to check a move is a sample — fifty records by hand, twenty attachments at random, or a statistically sized slice. The guides say something else as well: check everything when the data carries real risk. That is the half most firms skip. We compare every record and fingerprint every value on both sides, because a sample tells you the move was probably fine and a buyer who has just changed systems needs better than probably.
-
We will not call the report a certificate.
It states findings, not opinions, and it gives no assurance. A word that implies otherwise would be claiming something the document itself disclaims.
-
We will not touch anything on the health check.
The $390 tier reads and reports. Nothing is merged, deleted or edited. If you want changes made, that is the tier above, and it is a separate decision you make after seeing what is actually there.
How much we take on, and where the line is.
Cleaning covers one system, up to 25,000 records. Moving covers one source to one target, up to 25,000. Those caps are set from 117 real job postings that stated their volume: the median job in that set was 600 records, and 25,000 covers 85% of them. Above that we quote separately rather than pretend the same price still fits — a bigger job is more work, not the same work with a larger number attached to it. The monthly plan covers one system up to 100,000 records, which is a higher ceiling than the rest because it is a smaller job: it re-runs the same checks and nothing else. There is no move to correct and no mapping to agree.
The $390 check also tells you how many pairs you will have to decide on yourself before you commit to the tier above — the machine merges what it is sure of, and the rest is a list somebody at your end has to work through. You get that number before you commit to the work, not after.
Going through the results with you is one session. We walk you through the groups that need a person to decide, and anything still left is written into the report for your team to work through in their own time. We would rather tell you that now than discover it halfway through.
It drifts back. That is not a failure.
Cleaning a database once is like cleaning a kitchen once. The forms keep submitting, the imports keep arriving, and the same two salespeople keep adding the same person. Any firm that tells you a one-off clean holds is selling you the wrong thing.
So the monthly plan exists, and it buys exactly one thing: the same checks run again every month, with a report by the fifth working day, so you find out while it is small. It is optional, nothing switches off without it, and the work we did stays done.
Three promises, all three written down.
It reconciles, or you don’t pay.
If the report cannot be made to reconcile, the deposit is refunded. A guarantee with an asterisk is worse than no guarantee.
Late is our problem, not yours.
Miss the delivery date and 20% comes off the price, once. The clock starts when we have the complete access pack, and it pauses whenever we are waiting on you.
You own the findings, including the unflattering ones.
Every check names what it examined and what it could not. A report that only says what went well is a report nobody can act on.
Records drift back. Somebody has to keep checking.
Cleaning a database once is like cleaning a kitchen once. The same checks run again on a rhythm, and that is the same job as watching anything else we build — so it is one ladder, and you pick how much of it you want.
Watch
$549 per month
“It runs. I do not want to be the one who notices when it stops.”
- We are told the moment a run fails, and we tell you
- A written note every month of what ran, what failed and what we did
- Broken things fixed by the next working day
- Your accounts stay yours — we hold access, not ownership
Guard
$849 per month
“I need to know about the runs that finish and quietly do nothing.”
- Everything in Watch, plus:
- We check the runs that finish but do nothing, where there is something to count
- Broken things fixed the same working day
- Changes are unlimited — they join a queue and we work through it in order
Managed
$1,200 per month
“I do not want to run it at all. I want somebody whose job it is.”
- Everything in Guard, plus:
- We operate it, not you
- A named person, and the name of their backup
- Reachable outside business hours when something is genuinely down
- Your changes are worked ahead of everyone else’s
Monthly, in advance. Thirty days’ notice either way, from either side. Offered after a build, never before — we will not sell you a plan to watch something we have not seen.
Not sure which of these your records need?
Write to us and describe the problem in your own words — no form, no call. We will tell you which one we would pick, what is good and bad about that choice, where it would be the wrong buy, and what it costs. If the honest answer is that you do not need any of it, we will say that instead.
If the records in question are in a CRM rather than a spreadsheet, the HubSpot page says which fields we read, which count is the one to trust, and why the obvious duplicate is the one that platform cannot hold.
Start with the cheap one.
$390 buys the honest answer about what is in there, and it changes nothing. If it turns out your records are fine, we will tell you that and you will have spent $390 finding out instead of $2,950 fixing a problem you did not have. We reply within one business day.
You can get a free scan of your records from several firms, and they are worth running. A free one is the opening move of a sales conversation, so it has to find something. This one is paid for that reason: no fee here depends on what we find, and we did not build, migrate or operate the system we are reading.